Artificial Neural Networks for Material Identification, Mineralogy and Analytical Geochemistry Based on Laser-Induced Breakdown Spectroscopy

Artificial Neural Networks (ANN) are used nowadays in a broad range of areas such as pattern recognition, finances, data mining, battle scene analysis, process control, robotics, etc. Application of ANN in the field of spectroscopy has generated a long-standing interest of scientists, engineers and application specialists. The ANN’ capability of producing fast, reliable and accurate spectral data processing has become, in many cases, a bridging mechanism between science and application. A particular example of how ANN can transform plasma emission spectroscopy, that is quit challenging to model, into a turnkey ready to use device is described in this Chapter. Laser-Induced Breakdown Spectroscopy (LIBS) is a material-composition analytical technique gaining increased interest last decade in various application fields, such as geology, metallurgy, pharmaceutical, bio-medical, environmental, industrial process control and others (Cremer & Radziemski, 2006; Miziolek et al., 2006). It is in essence a spectroscopic analysis of light emitted by the hot plasma created on a sample by the laser-induced breakdown. LIBS offers numerous advantages as compared to the standard elemental analysis techniques (X-ray fluorescence or X-ray diffraction spectroscopy, inductively coupled plasma spectroscopy, etc.), such as: capability of remote analysis in the field, compact instrumentation, detection of all elements and high spatial resolution. Such features as minimum or no sample preparation requirement and dust mitigation using “cleaning“ laser shots are especially important for field geology and remotely operated rover-based instruments. As result, LIBS instruments have been selected as payloads for the 2011 Mars Science Laboratory mission led by the National Aeronautics and Space Administration (Lanza et al., 2010) and the ExoMars mission on Mars planned for 2018 and led by European Space Agency (Escudero-Sanz et al., 2008). Despite of the advantages, the main challenge is still the retrieval of accurate information from measured spectra. LIBS spectral signals, composed mostly of narrow emission lines, are complex nonlinear functions of concentrations of measured constituents and instrument